1. Field of the Invention
The invention relates to a method for the acquisition of a measurement data set for a respirating examination subject by means of magnetic resonance technology, and a magnetic resonance apparatus, and a computer-readable data storage medium for implementing such a method.
2. Description of the Prior Art
Magnetic resonance (MR) is a known technology with which images from the interior of an examination subject can be generated. Expressed simply, the examination subject is placed in a magnetic resonance imaging scanner, in a strong, static, homogenous base magnetic field, also called a B0 field, having a field strength of 0.2 tesla-7 tesla and more, such that the nuclear spins of the subject orient themselves along the base magnetic field. In order to trigger magnetic resonance signals, the examination subject is irradiated with high frequency excitation pulses (RF pulses), the triggered magnetic resonance signals are detected and entered into a memory that represents a mathematical domain known as k-space, and MR images are reconstructed on the basis of the k-space data, or spectroscopy data are determined. For the spatial encoding of the measurement data, rapidly activated magnetic gradient fields are superimposed on the base magnetic field. The recorded measurement data are digitized and stored as complex number values in a k-space matrix. From the k-space matrix populated with data values in this manner, an associated MR image can be reconstructed, for example, by means of a multi-dimensional Fourier transformation.
The respiratory movement of a patient that is to be examined by means of MR can lead to so-called ghosting, to blurring, and/or to intensity losses in the images generated, as well as registration errors between generated images particularly in an examination of the organs of the thorax and the abdomen, i.e. of examination regions affected by respiratory movement. These artifacts can make it difficult for a physician to perform an analysis on the basis of the images, and can lead to lesions being overlooked, for example. Numerous techniques exist in the prior art for reducing artifacts resulting from respiratory movement. One of these techniques is respiratory gating. Respiratory gating is a technique with which, during the MR measurement, the respiration of the patient is recorded and assigned to the acquired measurement data. With respiratory gating, only measurement data are then used for reconstruction for which the associated recorded respiratory movement fulfills certain specifiable criteria.
The respiration of the patient can be detected with external sensors, e.g. a pneumatic cuff or belt, or with MR signals, so-called navigators. A navigator is normally a short sequence of the MR signals, e.g. acquired from the diaphragm or other signal sources in the examination subject, the movement of which is correlated to the respiration of the patient. The respiratory movement can be reproduced from the position of the diaphragm or the other signal sources. The respiratory position is normally (but not necessarily) determined solely from the data recorded with two navigator sequences. The one navigator sequence is a reference measurement (normally the first navigator sequence), for the other navigator measurement, which is assigned to the determined respiration position.
In the following, reference shall be made to respiration triggers, with or without navigators. Respiration triggering is understood in this context as a technique that synchronizes the image generating MR measurement with the respiration of the freely respirating patients, and with which it is attempted to acquire a defined packet of measurement data during a distinctive phase of the respiratory cycle. If, in doing so, a specific layer is excited only once for each triggering, then the effective TR (the time period between successive excitations of a layer) of the sequence is the same, or is a multiple of the average respiratory cycle of the patient. In contrast, with respiratory gating, the repetition rate, particularly the TR thereof, is independent of the respiration of the patient. The repetition rate is controlled, rather, by means of a parameter, or by means of an additional physiological signal, e.g. an ECG.
For respiratory triggers with navigators, monitoring phases, during which the navigator sequence for recording the respiratory signal is repeated, alternate with measurement phases, during which the imaging sequence is carried out. During the monitoring phases, the imaging sequence is normally not carried out. In this respect, the temporal scanning rate of the respiratory movement is only limited in terms of a lower limit by the duration of the navigator sequence. The scanning rate is therefore normally higher than with respiratory gating.
If the respiratory signal is recorded using an external sensor, e.g. a pneumatic belt, then the scanning rate of the physiological signal is normally substantially higher than with the use of navigator measurements, such that digital filters and temporal averaging is normally possible. Furthermore, for a measurement of the physiological signal, the imaging measurement must be uninterrupted. Accordingly, with respiratory gating using an external signal, the acceptance window is not defined as a function of the respiratory position, but instead, is normally defined as a time window. The opening of the time window occurs, for example, when the measured physiological signal (which is a function of the chest measurement when a respiratory belt is used), falls below a threshold during the exhalation, and closes when this threshold is again exceeded during an inhalation (see, e.g. Craig E. Lewis et al. “Comparison of Respiratory Triggering and Gating Techniques for the Removal of Respiratory Artifacts in MR Imaging;” Radiology 1986; 160, Pages 803-810).
In the following, primarily respiratory gating methods, with a recording of the respiratory position by means of navigator measurement, shall be considered. For respiratory gating with navigators, the navigator sequence is interlaced with the imaging sequence, for example, and a diaphragm position measured using a navigator is subsequently assigned to the imaging data acquired directly thereafter (or before).
A distinction is made between retrospective and prospective respiratory gating.
With retrospective respiratory gating the respiratory movement is detected and recorded during the MR measurement, but not evaluated. Instead, the k-space that is to be recorded is measured repeatedly. For the reconstruction, only a portion of the measured data are referenced, preferably that data in which the respiratory signal lies within a specific window for a distinctive respiratory position. If a specific k-space data point that is necessary for the image reconstruction is repeatedly measured within the distinctive window, then the data can be averaged. If, instead, a data point is always measured outside of the window, then that data point deviating the least from the distinctive position can be used for the reconstruction.
With prospective respiratory gating, the physiological respiratory signal measured using a respiratory sensor (e.g. the diaphragm position measured with a navigator sequence) is evaluated during the measurement, and the MR measurement is controlled, based on the recorded physiological signal. In the simplest embodiment, the so-called acceptance/rejection algorithm (ARA), the measurement of an imaging data packet (and if applicable, the associated navigator sequence) is repeated until the physiological signal falls within a previously defined acceptance window.
One example of an acceptance/rejection algorithm of this type and, at the same time, the first description of respiratory gating with navigators, is described in the article by Todd S. Sachs, Craig H. Meyer, Bob S. Hu, Jim Kohli, Dwight G. Nishimura and Albert Macovski: “Real-Time Motion Detection in Spiral MRI Using Navigators,” MRM 32: Pages 639-645 (1994). The authors acquired one or more navigators for each excitation of a spiral sequence. The navigators were acquired here following the acquisition of the image data. Different navigators are distinguished by their spatial orientation. From each navigator, a spatial displacement along the axis of the navigator in relation to a reference navigator is calculated using a cross-correlation. The navigator scan acquired following the first imaging scan is used, in each case, as a reference. A specific imaging scan is repeated until the spatial displacement determined with the navigator, in relation to the reference, is less than a threshold value provided by a user. This, therefore, relates to an acceptance/rejection algorithm based on one or more spatial displacements.
Another example of an acceptance/rejection algorithm is described by Wang et al. in “Navigator-Echo-Based Real-Time Respiratory Gating and Triggering for Reduction of Respiratory Effects in Three-Dimensional Coronary MR Angiography,” Radiology 198; Pages 55-60 (1996). In this case, the physiological signal is the displacement of the diaphragm position, determined with a navigator, in relation to a reference state. One difference from the work by Sachs et al. is that, in each case, a navigator is acquired before and after the imaging scan, and that the imaging scan is then only accepted if the displacement determined by means of both navigators is less than the threshold value.
In order to determine the acceptance window, a so-called pre-scan is normally carried out for each patient, in which the respiratory movement is recorded, for example, with the navigator sequence, but imaging data are not yet acquired.
Prospective respiratory gating is normally more efficient than retrospective respiratory gating. A prerequisite for prospective respiratory gating is a real-time capability of the normally-provided control software for the MR apparatus. For this purpose, real-time capability means that data measured with the sequence (in this case, the sequence comprises imaging and navigator sequences) can be evaluated during the sequencing, and the further course of the sequencing can be influenced by the results of this evaluation, wherein the time period between recording the data and influencing the further course is short in comparison with the typical time constants of the respiratory movement (in this case, particularly, the respiratory cycle of a human being, which amounts to between 3 and 10 seconds).
The main problem with the acceptance/rejection algorithm is that the respiration of the patient frequently varies during the examination. The variations in the respiratory movement can be, thereby, such that the respiratory positions within the once specified acceptance window are rarely, or no longer, detected. This leads to extended acquisition periods and can even lead to the measurement not being completed at all in the normal manner.
A large number of alternative prospective gating algorithms exist, which either attempt to improve the efficiency with respect to the acceptance/rejection algorithm, or to make the measurements more robust, for the case that the respiration during the measurement varies or drifts. Common to all of this is that—as long as the respiration is recorded with a navigator—the prospective decision is based solely on the last measured respiration position (e.g. the last measured diaphragm position or the last measured diaphragm positions).
The most important algorithm, by far, that addresses this problem is “Phase Ordering With Automatic Window Selection” (PAWS), which is described, for example, in the article by P. Jhooti, P. D. Gatehouse, J. Keegan, N. H. Bunce, A. M. Taylor, and D. N. Firmin, “Phase Ordering With Automatic Window Selection (PAWS): A Novel Motion-Resistant Technique for 3D Coronary Imaging,” Magnetic Resonance in Medicine 43, Pages 470-480 (2000) and in the US patent, U.S. Pat. No. 7,039,451 B1. PAWS finds a final acceptance window during the runtime, and can thus react in a flexible manner to a changing respiration. A further goal of PAWS is to ensure a certain degree of “phase-encode ordering” (or in short, “phase ordering”). This means that adjacent lines in the k-space are acquired in similar respiration states. In particular, a variation in the respiratory state during acquisitions in the vicinity of the k-space center, which is particularly sensitive to movement, is to be avoided. PAWS was developed for a 3D Cartesian acquisition technique. The ky-kz array system used for this acquires a complete kx-kz plane of the 3-dimensional k-space following each navigator. The modulation of the k-space signal along the kz axis resulting from the transcendental state after interrupting the stationary steady state by the navigator (as well as potential activated preparation pulses, or the waiting for a further physiological signal, such as an EKG trigger) on the kx-kz plane, is therefore smooth. Discontinuations may arise in the ky axis as a result of residual movement, which can be manifested in the image as artifacts and blurring along the first phase encoding axis ky. This does not only apply when the segment border exists in the vicinity of the k-space center. Peristaltic movements, as well, which are not detected by the respiratory sensor, can lead to artifacts in the images.
PAWS exists in different so-called “bin” variations. In PAWS, the width of the final acceptance window is determined. The respiratory positions encompassed by this acceptance window are found automatically during the runtime, in contrast to acceptance/rejection algorithms. The filling of k-space occurs in clusters. A cluster (in the original work, the term “bin” is used instead of “cluster”) is characterized by a respiratory position range, an acceptance range, and comprises all k-space lines that have already been measured, after which a respiratory position in the respiratory position range assigned to the cluster is measured. In the n-bin variation of PAWS a respiratory position range is covered by n successive clusters, the width of which range is the same as the acceptance window.
Furthermore, a starting position in the k-space is assigned to each cluster, wherein the number of different starting positions is n. Different starting positions are assigned to clusters with adjacent respiratory positions where n>1. As soon as a respiratory position assigned to a cluster is measured with the navigator, the measurement of a k-space line that has not yet been measured within said cluster is initiated. The decision regarding which k-space lines still to be measured are selected takes into consideration, as a whole, the already acquired k-space lines of adjacent clusters as well. By way of example, a still missing k-space line is selected such that an arbitrary group of n adjacent clusters is complete to the greatest degree possible, wherein the arbitrary group of n adjacent clusters contains the cluster to which the current measured respiratory position is assigned; i.e. the group of n adjacent clusters comprising the largest possible number of different k-space lines. As soon as an arbitrary group of n adjacent clusters comprises all of the k-space lines that are to be measured, the measurement is stopped, because the overall variation in the respiratory position is limited in these measurement data, thereby, to the acceptance window.
The n different starting points and clusters of the n-bin variation of PAWS normally result in n segments in the k-space. For this, each segment consists of adjacent k-space lines. The variations to the respiratory positions within a segment measured with the navigator correspond to the position range assigned to a cluster (in the original work, the term “bin size” is used), and thus one nth of the acquisition window. The variation to the respiratory position is greater over the course of the entire k-space, and has an upper limit as a result of the specified acceptance window. The lines belonging to the same segment are measured during similar respiratory states. Thus, the modulation of the signal changes with the respiration at the segment borders. As a result, position jumps occur at the segment borders. An aim of the different bin-variations of PAWS is to displace the segment borders away from the movement sensitive k-space center. Another aim is to obtain a greater degree of efficiency.
In the previously mentioned article by Jhooti et al., as well as in the follow-up work by P. Jhooti, P. Gatehouse J. Keegan, A. Stemmer, D. Firmin: “Phase ordering with Automatic Window Selection (PAWS) with Half Fourier for Increased Scan Efficiency and Image Quality;” Proc. Intl. Soc. Mag. Reson. Med. 11 (2004); Page 2146, the 1-bin, 2-bin, 3-bin, and 4-bin variations are compared with one another. The result of this comparison shows that the 1-bin and the 2-bin variations of PAWS are the most efficient, i.e. for a given width of the acceptance window, the measurements are completed most quickly. The 1-bin variation is discarded because it does not allow for “phase ordering,” the 4-bin variation (and higher) is discarded due to lower efficiency. The 3-bin variation is less efficient than the 2-bin variation. The reason for this is the unidirectional growth direction of the cluster with starting positions at the left and right k-space edges. As soon as the gap between one of these peripheral clusters and the central cluster (with a starting position in the k-space center, and a bidirectional growth direction) is closed, then said clusters continue to grow until the gap between the other peripheral clusters and the central cluster is closed, as soon as a respiratory position is measured that is assigned to the first peripheral cluster. This normally leads to multiple k-space lines acquired at the cluster borders (segment borders). This problem does not exist with the 2-bin variation. In this variation, every second cluster grows in a unidirectional manner from the left-hand k-space edge, through the k-space center, toward the right-hand k-space edge, and the remaining clusters grow in a unidirectional manner from the right-hand k-space edge, through the k-space center, toward the left-hand k-space edge. The measurement is complete as soon as two adjacent clusters (with opposite growth directions) “meet.” However, with a symmetrical scanning of the k-space, as is the case with the 2-bin variation, the cluster border frequently lies in the vicinity of the k-space center, which is particularly sensitive to movement, which may lead to strong image artifacts. The probability of cluster borders lying in the vicinity of the k-space center is substantially lower with the use of partial Fourier (i.e. an asymmetric scanning of the k-space).
Of practical relevance, therefore, are the so-called 2-bin and 3-bin versions of PAWS, wherein, with symmetrical scanning, the 3-bin variation is preferred, and with asymmetric scanning, the 2-bin variation is preferred. This analysis is based on a 2-bin variation, in which the starting position alternates between the left-hand and right-hand k-space edges of adjacent clusters. Accordingly, the clusters grow, respectively, from the starting positions assigned thereto, firstly toward the k-space center.
It should also be noted that in some works, only a single respiratory position is assigned to a cluster. The width of the final acceptance window then amounts to n times the resolution of the respiratory signal. In this alternative formulation, one obtains a flexible selection of the acceptance window in that one first enlarges the respiratory position measured with the sensor, such that the n adjacent resulting respiratory positions cover a respiratory range corresponding to the width of the acceptance window.
PAWS was originally developed for a ky-kz array system, with which, in each case after recording the respiratory signal, all k-space lines are acquired with a specific value of the second phase-encoding gradient (in the kz axis). Accordingly, the “phase-ordering” is also limited to a Cartesian k-space axis, which can lead to a higher occurrence of residual movement artifacts in this axis.
Another prospective gating algorithm, which addresses the problem of a varying respiration, is the Diminishing Variance Algorithm (DVA) (Todd S. Sachs, Craig H. Meyer, Pablo Irarrazabal, Bob S. Hu, Dwight G. Nishimura: “The Diminishing Variance Algorithm for Real-Time Reduction of Motion Artifacts in MRI;” MRM 34; Pages 412-422 (1995)). First, in an initial phase, the k-space that is to be acquired is completely recorded, without gating, and for each acquisition period, the displacement of the respiration position in relation to a reference position is measured with a navigator, and recorded. At the end of the initial phase, the most frequent respiratory position is determined using a histogram, and designated as the mode for the statistical distribution. The k-space data, the stored respiratory positions of which deviate the most from the mode, are then re-acquired together with the navigator, and the histogram (and thereby the mode) is updated. This re-acquisition of the k-space data, which in each case deviate the most from the current mode, is continued until all of the respiratory positions lie within an acceptance window of a given width, or a time limit has been reached.